[riscv64] disable vectorizer during LTO

With LTO, vectorization needs to be disabled separately in the linker flags as well.

Change-Id: I71d813974a9b7a40954d3dab5b1704baed610fd0
diff --git a/cc/config/riscv64_device.go b/cc/config/riscv64_device.go
index bcd0060..f76e8c9 100644
--- a/cc/config/riscv64_device.go
+++ b/cc/config/riscv64_device.go
@@ -43,6 +43,8 @@
 		// Equivalent to "-munaligned-access", but our clang doesn't have that yet.
 		"-Xclang -target-feature -Xclang +unaligned-scalar-mem",
 		"-Xclang -target-feature -Xclang +unaligned-vector-mem",
+		// Until https://gitlab.com/qemu-project/qemu/-/issues/1976 is fixed...
+		"-Wl,-mllvm,-vectorize-loops=false",
 		// We should change the default for this in clang, but for now...
 		// (https://github.com/google/android-riscv64/issues/124)
 		"-Wl,-mllvm -Wl,-jump-is-expensive=false",